Measuring similarity in feature space of knowledge entailed by two separate rule sets

This paper addresses the task of comparing two rule sets induced within the same feature space for measuring the knowledge entailed jointly by the two. A procedure that quantifies the similarity of knowledge entailed by two separate rule sets in a given feature space is proposed. A formalized description of the proposed procedure along with its computational complexity analysis, applicability and utility is presented. Application of the proposed procedure is demonstrated using two rule sets from the computer security domain.